Drescher, A. & Detournay, E. (1993). GCotechnique 43, No.

3,443-456

Limit load in translational failure mechanisms for associative and

non-associative materials

A. DRESCHER* and E. DETOURNAY?

Translational failure mechanisms for estimation of Dans cet article, on s’intkesse aux mkchanismes
the limit load in geotechnical boundary-value prob- de rupture par translation dans le but d’estimer la
lems are considered. The equivalence of the energy charge limite de probkmes gkotechniques.
balance approach and the equilibrium of forces L’kquivalence entre I’approche i?nergCtique et la
approach for limit load evaluation in this type of mkthode par kquilibre des forces est Ctudike en
failure mechanism is studied in detail. The soils are d&ail pour ce type de mechanismes de rupture. Le
modelled as rigid-perfectly plastic materials with sol est assimili? g un materiau rigide parfaitement
flow rules associative and non-associative to a plastique, characterisi! par une rkgle d’kcoulement
linear Mobr-Coulomb yield condition. It is demon- qui est associke (matkriau standard) ou non-
strated that for non-associative materials with a associke (matkriau non-standard) au critkre lin-
coaxial flow rule, the limit load in translational taire de Mohr-Coulomb. Pour un mattriau
failure mechanisms corresponds to a fully devel- non-standard on dkmontre que la charge limite
oped residual strength due to non-associativeness, pour un mkchanisme de rupture par translation
and is lower than the limit load for the associative correspond en fait $ une rksistance residuelle like P
material. However, the limit load for non- la non-associativitk de la rkgle d’kcoulement; de
associative materials with a non-coaxial flow rule plus la charge limite est infirieure $ celle pour un
may be the same as for the associative material. materiau standard. Par centre, si la rt+gle
Implications for the interpretation of one of the d’koulement n’est pas coaxiale, la charge limite
limit equilibrium methods are discussed. pour un materiau non-standard est kquivalente A
celle pour un materiau standard. Finalement, on
discute les implications de cette analyse pour
I’interpretation d’une des mkthodes d’kquilibre
KEYWORDS: failure; limit state analysis; plasticity. limite.

INTRODUCTION allows analytical and straightforward calculation

This Paper deals with a classic geotechnical of the limit load. In addition, solutions for this
problem, namely the limit load that an engineer- model serve as references for the verification of
ing structure can exert on, or receives from, a numerical solutions, which use much more
mass of soil. Usually, the notion of the limit load complex constitutive responses.
is identified with impending collapse; in this Within the concept of the rigid-perfectly plastic
Paper it is taken in a broader sense, being inter- model, two groups of analytical methods for limit
preted also as the load at which the collapse may load evaluation can be identified: limit stress and
actually take place. limit equilibrium methods, and limit analysis
There is a vast technical literature dealing with methods. The first use only the yield condition;
limit loads for particular structures such as the second also involve the associative flow rule
retaining walls, foundations, anchors and tunnels, and provide useful bounds to the limit load.
as well as with fundamentals of prediction These bounds follow from uniqueness of the limit
methods. The assumption underlying most load for the associative material. Real soils, when
methods of limit load prediction is the rigid- modelled as rigid-perfectly plastic solids, do not
perfectly plastic response of soils. This usually obey the associative flow rule and the limit load is
no longer unique. The only available information
is that a non-associative material cannot be
Discussion on this Paper closes 3 January 1994; for stronger than the associative one. This implies
further details see p. ii. that limit analysis may overestimate or underesti-
* Department of Civil and Mineral Engineering, Uni- mate the limit load, depending on the problem
versity of Minnesota. considered (passive or active) and the method
t Schlumberger Cambridge Research, UK. applied (kinematic or static).

44.3
444 DRESCHER AND DETOURNAY

No analytical methods have been developed for sands and clays. The various types of non-
non-associative materials that parallel the limit associative material proposed in the literature are
analysis methods for associative materials. presented, and the corresponding limit loads
Numerical (finite element method) computations obtained from the kinematic method are
have been performed instead to provide informa- analysed. Implications for the interpretation of
tion on the difference in the limit load for the two one of the limit equilibrium methods are dis-
types of material (Davis & Booker, 1973; Zienkie- cussed.
wicz, Humpheson & Lewis, 1975; Grifflths, 1982;
Mizuno & Chen, 1983; de Borst & Vermeer,
1984; Vermeer & de Borst, 1984; Famiyesin,
1990). The results of these studies suggest that in KINEMATIC METHOD OF LIMIT ANALYSIS
some boundary-value problems, for example the IN APPLICATION TO TRANSLATIONAL
bearing capacity problem, the difference in the FAILURE MECHANISMS
limit load for the non-associative coaxial material According to the limit analysis theorems, the
and for the associative material should be small. kinematic method provides an upper bound on
This Paper re-examines translational failure the true limit load inducing collapse, and a lower
mechanisms as engineering approximations of bound on the true limit load resisting collapse
true collapse mechanisms for the evaluation of (reactions). The limit load is determined by equat-
limit loads for various kinds of rigid-perfectly ing the rate of internal work that is dissipated in
plastic materials. The approximation involves the a kinematically admissible failure mechanism
simplification of the mechanism to a system of (internal rate of energy dissipation) and the rate
regions undergoing rigid-body translation, and of work of external forces (see Davis, 1968; Chen,
separated by thin transition layers in which the 1975). If considerations are restricted to two-
material is experiencing shear distortion and dimensional problems, this equality, often termed
possibly dilation. This type of failure mechanism ‘the energy balance equation’, can be expressed
has been used for the derivation of limit loads for mathematically as
associative materials, within the framework of the
kinematic method of limit analysis. In this Paper,
translational failure mechanisms are also applied
to non-associative materials, in an attempt to
predict limit loads for these materials analytically.
=~~i~idS+~~iuidA (1)
There are several reasons for focusing attention
on translational failure mechanisms. First, the
analytical effort is greatly reduced, since no defor-
where the first two terms represent the rate of
mation takes place within regions undergoing
work done by the stresses oij over the deforma-
rigid translation. Second, thin shear layers
tion rates D, and dissipated within region A, and
separating rigid regions are often observed at col-
done by the tractions ti over the velocity jumps
lapse of soil and rock masses. Third, these layers
[uli and dissipated along the velocity discontin-
are modelled mathematically as velocity discon-
uity line L. The last two terms represent the rate
tinuity lines; this in turn leads to the equivalence
of external work of tractions ti over velocities vi
of the energy balance in the mechanism and the
at boundary S, and of body forces yi over velo-
equilibrium of forces acting on the rigid regions.
cities ui in region A. The stresses satisfy the yield
This equivalence plays an important role in the
condition and are related to the deformation
derivation of limit loads for non-associative
rates, resulting from a kinematically admissible
materials. Fourth, the response of a layer of a
velocity field, by the associative flow rule.
non-associative, coaxial, elastic-perfectly plastic
However, equation (1) cannot be interpreted gen-
material subject to shear has been studied pre-
erally as the principle of virtual work, as the
viously (Vermeer & de Borst, 1984; Vermeer,
stresses crij do not necessarily satisfy the local
1990), and may serve as the basis for interpreta-
equilibrium equations and boundary conditions.
tion of the limit load for non-associative, coaxial,
Consider now a particular class of failure
rigid-perfectly plastic materials.
mechanisms in which the failing mass is sub-
The main elements of the kinematic method of
divided into rigid regions undergoing translation,
limit analysis in application to translational
with straight boundaries identified as the velocity
failure mechanisms are summarized below. The
discontinuity lines. Since in rigid regions D, = 0,
equivalence of the energy balance approach and
equation (1) reduces to
the equilibrium of forces approach is studied in
detail and illustrated with examples. The yielding
of materials is described by a linear Mohr-
~ti~~]~dr.=~iia,dS+~7i”idA (2)
Coulomb yield condition, as usually assumed for
 TRANSLATIONAL FAILURE MECHANISMS 445

The dot product ti[uli represents the specific rate used for the case of frictional boundary condi-
of energy dissipation at the velocity discontinuity tions or non-associativeness, by selection of a
line, which, in terms of components normal and virtual velocity field that is not constrained by the
tangential to the velocity discontinuity line, can flow rule.
be expressed as (Fig. l(a)) Most limit load solutions in the literature for
frictional/cohesive materials use equation (2) (see
tiC”li= OtC”l*
- 0nCvln (3) Chen, 1975; Chen & Liu, 1990). Applications of
The components u, and un satisfy the yield condi- equation (4) were given by Mroz & Drescher
tion F(a,, a,) = 0 and are related to Cull and [u]” (1969) and Collins (1973); more recent results are
by the associative flow rule, which requires the given by, for example, Kolymbas (1987), Lin &
velocity jump vector to be normal to the yield Drescher (1989) and Drescher (1991).
condition in the Mohr diagram. In addition, since
the velocity discontinuity lines are straight, the
components [u], and [u]” are constant along LIMIT LOAD FOR ASSOCIATIVE MATERIALS
these lines. The evaluation of the limit load for materials
An important feature of translational failure with the linear MohrCoulomb yield condition
mechanisms is the possibility of expressing equa- has been studied extensively. Here, the main ele-
tion (2) as (see Appendix 1) ments of the analysis are discussed for further ref-
erence. Also, some non-classic examples are con-
sidered in order to illustrate possible extensions
(4) of the kinematic method.
For the linear Mohr-Coulomb yield condition,
which is a set of equilibrium equations of forces; with constant friction angle 4 and cohesion c, the
equation (4) applies to individual regions as well components 0,) cm, [u], and [u], satisfy (Fig. l(b))
as to the whole failure mechanism. This means
that if the problem is statically determinate the 0, = (T, tan 4 + c (5)
limit load can always be obtained by using either
equation (2) or equation (4). It also implies that
Culn= Culttan 4 (6)
for translational failure mechanisms, the energy Equation (6) means that the velocity jump is
balance equation (2) can be interpreted as an inclined to the velocity discontinuity at friction
expression of the virtual rate of work principle. angle 4. On substitution of equations (5) and (6)
The consequences of this interpretation of equa- into equation (3), a well-known result (Davis,
tion (2) are elaborated below, and it is shown that 1968; Chen, 1975) is obtained
an ‘apparent’ energy balance equation can be
ti[uli = c[v] cos & (7)
where [u] is the modulus of [uJ, and the specific
F\,WI" energy dissipation is independent of tractions. If
‘\

-J”
this result is used in equation (2), the standard
Velocity
discontiwty form of the energy balance equation is obtained
[VII
‘l
0” ut
ccos~[u,dL=~tiuidS+~yiuidA (8)
sL
(4 For example, for the problem of a smooth
retaining wall, passive state and two-triangle
failure mechanism (Fig. 2(a)), the energy balance
equation becomes

= Pu, - w, VI”- Iv*u*” (9)

where W is the weight of a region, L is the length
of the velocity discontinuity line, P is the limit
force, the subscript v signifies the vertical com-
ponent, and the velocities and their jumps result
from the hodograph (Fig. 2(b)). The least limit
(b) force is obtained by minimization of P with
Fig. 1. Traction and velocity jump (a) at a velocity dis- respect to angles 8.
continuity line; (b) for a linear Mohr-Coulomb yield con- Force P can also be obtained from the equi-
dition and associative flow rule librium of forces acting on the region, as the
446 DRESCHER AND DETOURNAY

(b)

Fig. 2. (a) Failure mechanism for a smooth retaining

wall; (b) hodograph, (c) polygon of forces (b) (cl

Fig. 3. (a) Failure mechanism for a rough retaining wall;

(b) hodograph; (c) polygon of forces
problem is statically determinate. Indeed, there
are only two unknown forces for each region, and
the weights of the triangles are known, as are the the same limit force is that the energy balance
cohesive components and the inclination of the equation is also, in the case of translational
frictional components of the forces acting at the failure mechanisms, an expression of the virtual
boundaries of each triangle; this results from the rate of work principle. The load deduced from the
integration of equation (5). A convenient way of energy balance equation is therefore independent
finding the limit force P is to use the polygon of of the particular choice of a kinematically admis-
forces, which is constructed beginning with region sible velocity field. An alternative explanation is
1 (Fig. 2(c)). given in the discussion of non-associative
Consider now the case when the wall is rough, materials.
with friction angle &,, and energy is dissipated The balance of energy approach and the equi-
along the wall (Fig. 3(a)). Two ways of incorpo- librium of forces approach can also be used for
rating wall friction in the energy balance equation walls comprising n differently inclined sectors. In
can be found in the literature, and lead to identi- the first approach, however, because the inclina-
cal force P. In the first (Chen, 1975), the energy tion of the total force on the wall is not known,
dissipation along the wall is calculated as the two energy balance equations must be used to
product of component P, = P sin 4, and the determine the vertical and horizontal components
velocity jump tangential to the wall equal to uZV of the total force. These equations are obtained
(Fig. 2(b)), with the resulting energy balance equa- by considering two different inclinations of the
tion velocity ue. Likewise, to determine the forces on
each sector of the wall, n energy balance equa-
tions corresponding to n inclinations of the veloc-
+ P,vzv = P” UCJ- w, VI” - w, 02” (10) ity u. must be used.
This can be illustrated by means of the example
In the second (Mroz & Drescher, 1969; Collins,
shown in Fig. 4 (n = 2, smooth wall, cohesionless
1969, 1973), the inclination of the velocity jump
material), for which the corresponding set of
across the wall is taken as &,, which leads to
energy balance equations is
energy dissipation equal to zero along the wall.
However, the velocities and velocity jumps used (P, cos a + P,)o, - W,uly - W2v2v = 0 (lla)
in the energy balance equation are now deter-
mined from the new hodograph shown in Fig. [PI cos (a + 0) + P, cos o]uO - WluwIv
3(b). As shown by Collins (1969), the force P so - Wzuwzv= 0 (tlb)
obtained is a rigorous upper bound to the true
limit force in the presence of interface friction. The resulting expressions for P,, P, , P and angle
The fundamental reason the two methods lead to 1 are given in Appendix 2. Fig. 5 shows the
 TRANSLATIONAL FAILURE MECHANISMS 447

4-

3-
2.

- 10”

Fig. 5. Broken smooth retaining wall: variation of non-

(b)

It is sometimes advantageous to use the equi-

librium of forces rather than the energy balance
approach, because it yields additional informa-
tion. Consider, for example, the problem of dis-
charge of granular materials from hoppers (Mroz
& Drescher, 1969; Drescher, 1991). Fig. 6(a)
shows a hopper with the failure mechanism con-
sisting of a number of geometrically similar
regions. The limit load in this problem is the force
supporting the material from below, and the kine-
matic solution provides the lower bound on this
force (reaction). In contrast to equation (8) which
gives the magnitude only of the limit force, the
equilibrium of forces approach also provides
Fig. 4. (a) Failure mechanism for a broken smooth
information on the magnitude of forces acting
retaining wall; (b) hodographs; (c) polygon of forces
along each sector of the wall. Fig. 6(c) shows the
polygon of forces for rough walls. Division of
dependence of the least non-dimensional force these forces by the length of the sectors gives an
PIYH’, angle 1 and the corresponding forces approximate distribution of wall stresses. As
P&H2 and P,fyH’ on angle E for 4 = 30” and shown by Mroz & Drescher (1969) and Drescher
H/h = 2. (1991), the stress distribution corresponding to

(4 (b) (C)

Fig. 6. (a) Failure mechanism for a rough hopper; (b) hodograph; (c) polygon of forces
448 DRESCHER AND DETOURNAY

maximum reaction compares favourably with the sector. This implies that the angle $ may vary
exact radial stress solution (Jenike, 1964). within the range 4 > 9 2 0. The angle $ is the
dilatancy angle, and $ = 0 means incompressible
LIMIT LOAD FOR NON-ASSOCIATIVE plastic flow. The end-points of the corresponding
MATERIALS tractions are also located along sector AB. The
This section extends the methodology of the non-associative non-coaxial flow rules, except de
kinematic method of limit analysis for determi- Josselin de Jong’s are represented by the velocity
nation of the limit load in translational mecha- jumps located at point A, but inclined at an angle
nisms for non-associative materials. This is ti whose range is given above. In all cases the
possible because for translational mechanisms end-point of the traction vector coincides with
both the energy balance approach and the equi- point A. De Josselin de Jong’s flow rule, in which
librium of forces approach apply to any rigid- $ = 0, allows for two locations of the velocity
perfectly plastic material, regardless of its flow jump vector, A and C, and two tractions.
rule (see Appendix 1). Various non-associative
flow rules are considered, and their influence on Materials with non-associative coaxialflow rules
the limit load is illustrated with examples. Next, For non-associative materials with a coaxial
the interpretation of the corresponding limit flow rule and a linear Mohr-Coulomb yield con-
loads is presented. dition, the components of tractions and velocity
jumps at the velocity discontinuity lines satisfy
Non-associativejlow rules (Fig. 8) (Davis, 1968)
Two types of non-associative flow rule are pro-
crt = 0, tan $* + c* (12)
posed in the literature. In the first, the assumption
of coaxiality of the principal directions of stresses Cvln= Cvl,tan ti (13)
and deformation rates, inherent in the associative
where
flow rule, is retained (Hill, 1956; Radenkovic
1961; Roscoe, 1970); in the second, non- tan $* = r] tan 4 (14)
coincidence of these directions is allowed for (de
c* = t/c (15)
Josselin de Jong, 1959; Spencer, 1964; Mandl &
Fernandez Luque, 1970; Mehrabadi & Cowin, cos $ cos 4
1978). As only failure mechanisms in which (16)
’ = 1 - sin ti sin 4
energy is dissipated along the velocity discontin-
uity lines are considered here, the difference The specific energy dissipation is then given by
between these flow rules can be represented in the ti[u], = [v] cos $[c* + a,(tan b* - tan $)I (17)
Mohr diagram. Fig. 7 shows a sector of the yield
condition with a stress circle tangent at point A, and depends not only on c, 4 and $ but also on
and crtrnax at point B. The associative flow rule is un , which is not known beforehand. This suggests
represented by the velocity jump [vii located at that the energy balance approach cannot be used
point A, and inclined at an angle $ = 4 to the effectively for a non-associative coaxial flow rule.
vertical. The tractions acting at the velocity dis- To show that this conclusion is incorrect, the
continuity line are given by vector OA. The non- equivalence of equations (2) and (4) is used. Like-
associative coaxial flow rules are represented by wise, for the associative flow rule, to determine
the velocity jumps located along the circular the limit load from equation (4) it suffices to
sector AB, with their orientation normal to that know the relation of 0, and 0, at the velocity dis-
continuity lines. This relation is given by equa-
tion (12), which can be regarded as a linear yield
condition with b* and c* replacing 4 and c. Since
the limit load depends on equation (12) and not
on the orientation of the velocity jump, a ‘ficti-
tious’ orientation and thus a fictitious flow rule
can be selected for which the specific energy dissi-
pation is independent of cr’,. This is fulfilled for
the fictitious flow rule associative to condition
(12), with the corresponding expression for the
specific energy dissipation given by
ti[vli = c*[v] cos (b* (18)
The use of a fictitious flow rule in conjunction
Fig. 7. Tractions and velocity jumps for coaxial and non- with the energy balance is legitimate, because the
coaxial non-associative flow rules energy balance equation (2) can also be inter-
 TRANSLATIONAL FAILURE MECHANISMS 449

Fig. 8. Tractions and velocity jumps for a linear Mohr-Coulomb yield con-
dition and a coaxial non-associative flow rule

preted as an expression of the virtual rate of work From equation (18) and the hodograph in Fig.
principle. With this interpretation, the velocity 9(b)
field is not necessarily constrained by the actual 2c* cos 4*
flow rule. (19)
H = 7 sin /3 cos (p + 4*)
Thus, the energy balance approach can be used
for a non-associative material with a coaxial flow which reaches the minimum for
rule if the specific energy dissipation and the
hodograph are taken as for the flow rule associa-
tive to the appropriately modified yield condition.
It should be stressed, however, that the failure
mechanism must be kinematically admissible for H,, = 4c*tan
the non-associative flow rule; this is satisfied if Y
the mechanism is statically determinate. An identical result is obtained from the polygon
To illustrate this conclusion, consider the of forces of Fig. 9(c). It can further be shown that
classic (statically determinate) problem of the equation (19) results from equation (17) and the
critical height of a vertical free-standing cut with hodograph shown in Fig. 9(d). In fact the energy
a single velocity discontinuity line (Fig. 9(a)). balance is now
cos +
- c*H + cos $(tan b* - tan Ic/)P”
cos p

= f yH2 tan fl cos (/? + $) (22)

The component P, derived from the polygon of

forces is
p = ceH tanB ~0s 4* (23)
(a) (b) ”
cm (B + d*)

n
which, when substituted into equation (22), gives
90” - fi - $3’ equation (19). Fig. 10 shows the dependence of
the non-dimensional critical height H,, y/c on the
W1 dilatancy angle $ for various values of 4.
Equation (21) differs from the equation given
c*L12 by Chen & Liu (1990)
V
(cl (d) H,, = 4c tan (24)
Y
Fig. 9. (a) Failure mechanism for a free-standing verti-
cal cut; (b) hodograph for a fictitious associative flow which implies that the critical height is indepen-
rule; (c) polygon of forces; (d) hodograph for a non- dent of the dilatancy angle. Chen & Liu (1990)
associative flow rule obtained equation (24) by assuming that the
450 DRESCHER AND DETOURNAY

parameters, yield conditions and boundary condi-

tions used by these authors, and the correspond-
ing non-dimensional bearing capacity q,/c; it also
gives results obtained from equation (26). The
numerical solutions yield values for qJc at the
onset of collapse higher than equation (26) for
Prandtl’s mechanism, and lower for Terzaghi’s
mechanism, irrespective of the value of +.
However, Prandtl’s mechanism should be regard-
ed as more adequate, as it gives the exact value of
the bearing capacity for $ = 4; the discrepancies
in results for $ = 4 can be attributed to numeri-
cal inaccuracies. For 4 = 20” and $ = o”, the dif-
ference between the numerical predictions and
equation (26) for Prandtl’s mechanism is about
12%. This is a significant difference: a possible
reason for it is given below.
4 1 I I I
Davis (1968), Davis & Booker (1973) and
0 15 30” 45” 60 Zienkiewicz, Humpheson & Lewis (1975) refer to
Cox’s (1963) static and kinematic solution based
Fig. 10. Free-standing vertical cut: variation of non-
on the method of characteristics as proof that the
dimensional critical height with dilatancy angle limit load in the bearing capacity problem should
be independent of the value of $. However, as is
demonstrated in Appendix 3, Cox’s solution leads
minimum value for H in equation (19) corre- to negative energy dissipation in the plastic
sponds to region and therefore, is thermodynamically inad-
missible.

Materials with non-associative non-coaxialflow

Consider now the more complex problem of rules
the bearing capacity of a strip footing on a For flow rules that lead to the velocity jump
weightless material, and two failure mechanisms located at the Mohr-Coulomb yield condition
attributed to Prandtl (1921) and Terzaghi (1943). (point A in Fig. 7), the dilatancy angle $ has no
Both involve a radial shear zone that can be effect on the limit load determined from a trans-
regarded as an infinite number of rigid regions lational failure mechanism. Indeed, the angle I,G
(Chen, 1975). The bearing capacity 4, for the affects only the hodograph; the yield condition
associative material is known and, when modified that governs the equilibrium of forces remains the
to incorporate 4* and c*, can be written as same. Although the specific energy dissipation
q, = c* cot d*(N, - 1) (26) depends on angle $ and component en similarly
to equation (17), the limit load can be determined
with the coefficient N, given by as for the associative material by use of equation
(7). This also pertains to de Josselin de Jong’s
NqP = exp (rr tan +*) tan* a+$ (27a) flow rule when the velocity jump is located at
( > point C. This time, however, equation (18) must
be used where
N~=~exp[2tan~*($-$)]
tan $* = p tan 4 (28)

;+$ c* = PC (29)
x set’ GW
( > cos*I$
where the superscripts P and T denote the (30)
’ = 1 + sin* 4
Prandtl and Terzaghi solutions respectively.
The bearing capacity values resulting from Equations (28) and (29) give much lower b* and
equation (26) are now compared with those for c* values than any other flow rule, and thus the
elastic-perfectly plastic associative and non- corresponding limit load will also be the least.
associative materials obtained numerically by The above analysis provides an alternative
Zienkiewicz, Humpheson & Lewis (1975), Mizuno explanation of the fact that the two methods of
& Chen (1983), de Borst & Vermeer (1984) and incorporating wall-soil interface friction in the
Famiyesin (1990). Table 1 gives the material retaining wall problem of Fig. 3 are identical. As
 TRANSLATIONAL FAILURE MECHANISMS 451

Table 1. Bearing capacity values: M-C = Mohr-Coulomb; D-P = Drucker-Prager matcb-

ing MohrXoulomb at plane strain; f-s = flexible, smooth; r-r = rigid, rough; rs = rigid,

v Yield 4 ti Foundation q,lc

condition

I I
20
0.3 M-C
20

f-s
r-r
f-s
r-r

de Borst & 40 40 78.1

4=
Vermeer (1984) 400 0.2 M-C r-s
40 20 73.7

Famiyesin 20 20 15.8
1153 0.3 DPP f-s
20 0 14.5

Equation (34) 20” 20” 14.8

I - M-C 40” 40 r-r or s 75.3
20” 0 13.0
40” 20 55.2

NqT
I T----t
A-_ M-C
20
40
20
40
20
40
0
20
r-r or s

This residual shear stress is the same as the

17.7
95.7
15.4
69.7

the tractions at the wall are the same in the two

methods, the method of Mroz & Drescher (1969) maximum shear stress of the stable response, and
and Collins (1969) is equivalent to modelling the hence is independent of the initial state. The ratio
interface as a layer of material with the flow rule of the residual shear stress to the stress normal to
associative to a linear Mohr-Coulomb yield con- the layer is given by
dition with friction angle 4,, whereas that of
Chen (1975) is equivalent to modelling the inter- cos * sin 4
: = tan 4* =
face as a layer of material with a non-associative n 1 - sin $ sin 4
non-coaxial flow rule and 1(1= 0. This type of
non-associativeness has no effect on the limit load The angle 4* represents the apparent residual
determined from a translational failure mechan- friction angle due to non-associativeness rather
ism. than to material softening, and has the same
value as the angle d* defined in equation (14).
Although Vermeer (1990) did not consider the
Interpretation of the limit load effect of cohesion, there is little doubt that equa-
To provide an interpretation of the limit load tion (15) would be recovered; c* can then be
for non-associative materials with coaxial flow regarded as the apparent residual cohesion.
rules, the results obtained by Vermeer (1990) are In a translational mechanism the real failure
used (see also Vermeer & de Borst, 1984). mode is approximated by a number of velocity
Vermeer considered shearing of a layer of a cohe- discontinuity lines regarded as limits of material
sionless elastic-perfectly plastic material with layers, whose behaviour can be modelled by the
I,$ < 4, subject to various non-isotropic initial above response. The limit load determined using
states of stress, and found that if the initial stress $r* and c* is then equivalent to the load when all
parallel to the layer is less than the perpendicular layers reach the residual strength, and this can be
stress, the shear stress-shear strain response is interpreted as the state of fully developed collapse
stable (hardening). If the initial stresses are of a non-associative material. Since the true
reversed, the response shows a peak beyond failure mechanism is more complex, and not all
which it becomes unstable (softening), with the shear layers reach the residual strength or fewer
shear stress eventually reaching a residual value. develop, the actual limit load may be higher.
452 DRESCHER AND DETOURNAY

Numerical errors notwithstanding, this may coaxial flow rule, for which the velocity jump
explain why the bearing capacity computed vector is located at C in Fig. 7. However, it is not
numerically for the onset of collapse is higher possible to determine the corresponding yield
than predicted by equation (26) and Prandtl’s condition and the magnitude of non-
mechanism. associativeness quantified by angle $ in the
The interpretation of the limit load for coaxial flow rule if they are not specified before-
materials with a non-coaxial flow rule, and the hand; there are only two available equations ((14)
end-point of traction located at point A in Fig. 7, and (15)) and three unknowns, 4, c and $ (or one
follows directly from two observations: the limit equation and two unknowns, 4 and Ic/).
load is identical with that for an associative
material, and the latter material cannot be
weaker than the former one. Thus, translational CONCLUDING REMARKS
failure mechanisms yield rigorous bounds to the In translational failure mechanisms the energy
true limit load for a material with this type of balance approach for evaluation of the limit load
non-coaxial flow rule; these bounds are the same is always equivalent to the equilibrium of forces
as for an associative material. For the end-point approach, because the energy balance equation
of traction located at point C, one may hypothe- can be interpreted as an expression of the virtual
size a similarity to a coaxial flow rule with &* rate of work principle. This observation has often
and c* now given by equations (28)-(30), i.e. the been made (Davis, 1968; Mroz & Drescher, 1969;
limit load is equivalent to the load when at all Collins, 1973; Justo, 1973; Goldscheider, 1988;
velocity discontinuities a ‘residual’ strength is Michalowski, 1989), although its significance has
reached. not generally been appreciated. The equivalence
of the two approaches not only allows verifica-
tion of the solutions, but also plays a key role in
LIMIT ANALYSIS AND LIMIT EQUILIBRIUM the derivations of limit loads for non-associative
METHODS materials. The equilibrium of forces approach
The relation of limit analysis methods and limit may also be advantageous in problems where the
equilibrium methods has been discussed pre- distribution of forces along rigid walls is of inter-
viously (Collins, 1973; Chen, 1975; Michalowski, est, for example the hopper problem. The only
1989). Here, some remarks on the method of limitation on the use of the equilibrium of forces
regions or blocks (Coulomb, 1773; Terzaghi, approach is that, in general, the failure mecha-
1943; Meyerhof, 1951; Enoki, Yagi, Yatabe & nism must be statically determinate, i.e. the
Ichimoto, 1991) are presented. Because this number of unknown forces for each region must
method operates with the global equilibrium of equal the number of equilibrium equations avail-
forces, and does not use the flow rule, the able.
resulting solutions are often regarded as approx- The limit load in translational failure mecha-
imate slip-line stress field solutions (Chen, 1975). nisms is governed by the relationship between the
However, since the solutions involve only a traction components at the velocity discontinuity
limited area, they cannot be identified as approx- lines. If this relationship is unaffected by the flow
imate limit analysis lower bound solutions, which rule, the limit load is also unaffected. This
require determination of the stresses in the whole explains the fact that limit loads for some non-
domain of a body. associative non-coaxial materials are the same as
From the equivalence of the energy balance those for an associative material: this type of
approach and the equilibrium of forces approach, non-associativeness does not affect the yield con-
it follows directly that any limit equilibrium solu- dition that governs the relationship of the trac-
tion is actually a limit analysis kinematic solution tion components at the velocity discontinuity
provided the inclination of forces at the bound- lines. Non-associative coaxial flow rules influence
aries of blocks is identical with the inclination of the limit load through a modified relationship of
tractions satisfying the yield condition; also, the the traction components. This also pertains to the
inclination of these forces must be compatible special case of de Josselin de Jong’s (1959) model.
with kinematically admissible relative motion of The limit load derived from a statically deter-
the blocks. These limit equilibrium solutions minate translational failure mechanism for a non-
provide an upper bound to loads inducing col- associative material with a coaxial flow rule can
lapse and a lower bound to loads resisting col- be interpreted as a load corresponding to a fully
lapse for a material with an associative flow rule. mobilized residual strength resulting from non-
If the inclination of forces does not satisfy the first associativeness (Vermeer & de Borst, 1984;
requirement above, the solution is equivalent to Vermeer, 1990). This limit load is lower than the
that for a material with a non-associative coaxial load for an associative material with the same
flow rule or de Josselin de Jong’s (1959) non- yield condition. Numerical solutions of the
 TRANSLATIONAL FAILURE MECHANISMS 453

bearing capacity problem and $ < 4 predict a velocity discontinuity line

limit load higher than the estimates presented for bearing capacity factor
Prandtl’s failure mechanism. This discrepancy can limit force
bearing capacity
be attributed to the fewer velocity discontinuities
boundary
in numerical solutions; the termination of the traction
numerical solutions at the onset of collapse and velocity
numerical inaccuracies may also contribute to modulus of [vii
overestimation of the limit load as seen for the velocity jump
associative material. The limit load corresponding component of [vii normal to L
to the residual strength is certainly a safe engin- component of [vii tangential to L
eering estimate of the true limit load. weight
The limit load obtained from a statically deter- angles
body forces
minate translational failure mechanism for a non-
Poisson’s ratio
associative, non-coaxial material other than a inclination of major principal stress to the
particular case of de Josselin de Jong’s (1959) x-axis
model is a rigorous bound to the true limit load. component of ti normal to L
In the excluded case, one is tempted to extend the component of ti tangential to L
findings for the coaxial flow rules, i.e. to assert stress
that the limit load corresponds to fully mobilized friction angle
‘residual’ strength. residual friction angle due to non-
The limit equilibrium solutions operating with associativeness
dilatancy angle
rigid regions are equivalent to limit analysis kine-
matic solutions, provided the inclination of forces
is derived from the yield condition and is compat-
ible with a kinematically admissible failure
mechanism. If the first requirement is not met, the
solutions correspond to non-associative materials APPENDIX 1. EQUILIBRIUM OF FORCES IN
with a coaxial flow rule of unknown (unless speci- TRANSLATIONAL MECHANISMS
fied beforehand) dilatancy angle $, or to one case Consider a two-dimensional rigid region A trans-
of de Josselin de Jong’s non-associative model. lating with velocity ui surrounded by a number of like
The analysis has been limited to two- reeions translating , 1... L’I”‘.
with velocities ui(‘), II.“’ L
dimensional translational failure mechanisms, but Asoss the bound;g lines of length L;, L, ,-. L,, ihe
applies equally to three-dimensional mechanisms. velocities undergo jumps [IJ]~“‘, [v]i”‘, [u]i’“‘, and
the tractions acting on the lines are ti(l), ti(‘), .__ti(n).
It can also be extended to materials with a non-
Fig. 11(a) shows the region as a triangle, although any
linear yield condition (rocks and frozen soils).
polygonal shape is acceptable; Fig. 1 l(b) shows the cor-
responding hodograph, where

o.(l) = 0, + [“],(I) (32a)

ACKNOWLEDGEMENTS 0,(2) = Vi + [“Ii(*) (32b)

A. Drescher wishes to express his gratitude to
“i(3) z lJi + [u]i(3) (32~)
the Carnegie Trust and the Royal Society for
financial support while on sabbatical leave at the
The energy balance equation for the triangle con-
University of Glasgow, where the work in this sidered is

s
Paper was initiated. Special thanks for support
are directed to Schlumberger Cambridge [u]i(” ti(l’ dL, + [u]i(” t,c’) dL, + [II]/“’
Research, where this Paper was completed. L, s I.1

x t/3) dL, = u.(l) ti(l) dL, + oi(21

s I.3 1.I L,

x t/2) dL, + I+‘~) tic3) dL, + vi yi dA (33)

NOTATION sA
s I.2 s LX
A region
c cohesion and, with the use of equation (32), can be written as
c* residual cohesion to non-associa-
tiveness 6i(~,lil’1dLI+Jj*ri(l)dL1
Dij deformation rate
G shear modulus
H, h height +L,+3)dL3+ly,dA)=0 (34)
H,, critical height
454 DRESCHER AND DETOURNAY

cos w
VW
‘” = ” tan (b2 - 4)

u0 1” = 0OZV+ Iu”,,l ~0s(a + B, - $1= u.

cos 0 sin (/?, - 4) + [cos 0 - tan (b2 - f#~)sin w]

xsinacos(a+/I-4)
x
tan (PZ - 4) sin (P1 - 4)

(36d)

p = w cos(a + B, - 4)
(374
1 1
sin(B, - 4)
P, =
K + w* + w cos(a+8,-44
tan(& - 4) ‘tan(8, - 44sinVI - 4)
x [sin a - cos a tan (p2 - $)] (37b)
P = J(P,Z + P,Z + 2P, P, cos d() (38)

P, sin a
I = tan-’ (39)
P, cos a + P,

APPENDIX 3. THERMODYNAMIC INADMISSIBILITY

(b) Cc) OF THE STRESS AND VELOCITY FIELDS IN THE
Fig. 11. (a) Tractions, velocities, and velocity jumps for BEARING CAPACITY PROBLEM FOR I/, < c#~
a rigidly translating region; (b) bodograph; (c) polygon Figure 12(a) shows the stress and velocity character-
of forces istics for the case of a rigid foundation resting on a
weightless frictional/cohesive soil with a non-associative
coaxial flow rule for 4 = 30” and 1(1= 0” (Cox, 1963;
Davis, 1968). The broken lines show the regions of
Since the velocity ui can be arbitrary, equation (34) leads
stress characteristics, and the full lines the correspond-
to
ing velocity characteristics (orthogonal). In regions ABF
and BGI the velocity characteristics are straight, and in
region BGF they form logarithmic spirals given by
r = r. es tanW/z) a characteristic (40a)
(35)
Equation (35) represents a set of global force equi- r = r,, e-’ cO”cp’Z’ p characteristic (40b)
librium equations for the region considered; this can be
shown graphically as a closed polygon of forces (Fig. with the origin of the r, Q co-ordinate system taken at
11(c)). A similar set applies to a three-dimensional point B, and 0 measured anticlockwise. Point B is not a
region. The derivation above uses the energy balance singular point for p-characteristics. The velocities satisfy
equation, and essentially follows that given by Micha- the equations
lowski (1989). The same result is obtained from the
principle of virtual rate of work applied to a rigid dv, + ug d[ = 0 a-characteristic (41a)
region undergoing translation (Mr6z & Drescher, 1969).
dub - U, dr = 0 /?-characteristic (41b)
Neither case uses the yield condition or flow rule, and
equations (33) and (35) apply to translational failure where V, and D#are the projections of the velocity vector
mechanisms for any rigid-perfectly plastic material. on to the a and B characteristics respectively and 5 is the
inclination of the major principal stress to the x-axis.
A Hill-type velocity field for a smooth foundation
APPENDIX 2. LIMIT LOAD FOR MECHANISM IN moving downwards with velocity uO” requires one
FIG. 4 velocity discontinuity line ACDGI. Since all points
1 along AB and ACD have the same velocity, region
“’ = ” tan (b2 - 4) ABJD moves as a rigid block. The known velocities
along DJ and DC1 formulate a well-posed character-
01” = 02” + C~IZI
cm (a + B, - $1 istic problem for the velocities in the region JDGI, and
can be found from equation (41).
sin (fit - f$) + sin a cos (a + /?, - I$)
“0 To prove that the corresponding velocity field will
(36b)
tan (& - 9) sin (B1 - 4) violate the thermodynamic requirement of non-negative
 TRANSLATIONAL FAILURE MECHANISMS 455

3
1

Circle
ti
D2

(W W)

Fig. 12. (a) Static and kinematic characteristics for the bearing capacity problem of an
incompressible weightless material; (b) bodograpb; (c) bodograpb for element D123; (d)
orientation of principal deformation rates in element 4567

energy dissipation, it suflices to find one point, or velocity than points 5 and 6 (Fig. 12(d)); it also applies
region, where to any point between points 4 and 5, and 6 and 7. The
orientation of the compressive principal deformation
bij D, < 0 (42) rate at any point within element 4567 is therefore verti-
Consider the element of the velocity characteristic mesh cal, whereas that of the tensile principal deformation
indicated in Fig. 12(a) as 4567. The velocities of points 5 rate is horizontal. As the only non-zero principal stress
and 6 are known, and those of points 4 and 7 can be is horizontal and compressive (uniaxial compression),
determined by construction of the hodograph for the the product of this stress and the horizontal deforma-
lowest row of elements, beginning from element D123 tion rate is negative, i.e. negative energy dissipation.
(Fig. 12(a)). Although a graphical construction of the This conclusion also applies to a rough foundation
hodograpb shown schematically in Fig. 12(b) may seem and the Prandtl-type mechanism with two velocity dis-
inaccurate, it su!Xces to determine the trend of the line continuity lines ACDGI and BCE... intersecting at
in the bodograpb representing points along 2347 with point C, and region ABC moving downwards as rigid.
respect to the circle representing points along D156. As In fact, the characteristic mesh remains the same, and
line D2 in Fig. 12(a) bends anticlockwise, line 23 in the the only difference in the hodograph pertains to the
bodograph turns left and away from the circle (Fig. magnitude of the velocity u,,‘, which now becomes twice
12(c)). This means that the velocity of point 3 is less uO” (Fig. 12(b)). A similar analysis can be performed for
than that of point 1, and that points 4 and 7 have lower 0 < $ < 4, and leads to the same result, i.e. negative
456 DRESCHER AND DETOURNAY

energy dissipation regardless of the magnitude of $: the Jenike, A. W. (1964). Steady gravity flow of frictional-
energy dissipation is positive only when $ = 4. cohesive solids in converging channels. J. Appl.
Mech. Trans. Am. Sot. Mech. Engrs 31, 5-l 1.
Justo, J. L. (1974). Discussion. Gtotechnique 24, No.1,
REFERENCES 106&107.
Chen, W. F. (1975). Limit analysis and soil plasticity. Kolymbas, D. (1987). Ice forces on conical offshore
Amsterdam: Elsevier. structures. Comp. Meth. Appl. Mech. Engng 6, 217-
Chen, W. F. & Liu, X. L. (1990). Limit analysis in soil 231.
mechanics. Amsterdam: Elsevier. Lin, M. & Drescher, A. (1989). Limit load in steady 3-D
Collins, I. F. (1969). The upper-bound theorem for rigid/ plastic flow around obstacles. Numerical models in
plastic solids to include Coulomb friction. J. Me& geomechanics: NUMOG III (eds S. Pietruszczak and
Phys. Solids 17, 323-338. G. N. Pande), pp. 297-304. London: Elsevier
Collins, I. F. (1973). A note on the interpretation of Applied Science.
Coulomb’s analysis of the thrust on a rough Mandl, G. & Fernandez Luque, R. (1970). Fully devel-
retaining wall in terms of the limit theorems of plas- oped plastic shear flow of granular materials.
ticity theory. Gkotechnique 24, No. 1, 106-108. GOotechnique 20, No. 3,277-307.
Coulomb, C. A. (1773). Essai sur une application des Mehrabadi, M. M. & Cowin, S. C. (1978). Initial planar
regles de maximis et minimis a quelques problemes deformation of dilatant granular materials. J. Mech.
de statique, relatif B l’architecture. Mem. Math. Phys. Phys. Solids 26, 269-284.
Acad. R. Sci. 7,343-382. Meyerhof, G. G. (1951). The ultimate bearing capacity
Cox, A. D. (1963). The use of non-associatedflow rules in of foundations. Giotechnique 2, No. 4, 301-332.
soil plasticity. Royal Armament Research and Devel- Michalowski, R. L. (1989). Three-dimensional analysis
opment Establishment, Report B, 2/63. of locally loaded slopes. Gkotechnique 39, No. 1,
Davis, E. H. (1968). Theories of plasticity and the failure 27-38.
of soil masses. In Soil mechanics: selected topics (ed. Mizuno, E. & Chen, W. F. (1983). Cap models for clay
I. K. Lee), pp. 341-380. London: Butterworth. strata to footing loads. Comp. Struct. 17, 51 l-528.
Davis, E. H. & Booker, J. R. (1973). Some adaptations Mr&, Z. & Drescher, A. (1969). Limit plasticity
of classical plasticity theory for soil stability prob- approach to some cases of flow of bulk solids. J.
lems. Proceedings of symposium on the role of plasti- Engng Ind. Trans. Am. Sot. Mech. Engrs 91, 357-
city in soil mechanics (ed. A. C. Palmer), pp. 24-41. 364.
Cambridge University Press. Prandtl, L. (1921). Uber die Harte plastischer Korper.
de Borst, R. & Vermeer, P. A. (1984). Possibilities and Nachr. kgl. Ges. Wiss. Gtittingen, Math. Phys.
limitations of finite elements for limit analysis, Klasse, 74-85.
GCotechnique 34, No. 2, 199-210. Radenkovic, D. (1961). TheoriC des charges limit&es,
de Josselin de Jong, G. (1959). Statics and kinematics in extension a la mecanique des soIs. Skmin. Plast.,
the failable zone of a granular material. Delft: Ecole Polytech., Publ. Sci. Techn. 116
Waltman. Roscoe, K. H. (1970). The influence of strains in soil
Drescher, A. (1991). Analytical methods in bin-load mechanics. Giotechnique 20, No. 2, 129-220.
analysis. Amsterdam: Elsevier. Spencer, A. J. M. (1964). A theory of the kinematics of
Enoki, M., Yagi, N., Yatabe, R. & Ichimoto, E. (1991). ideal soils under plane strain conditions. J. Me&
Generalized limit equilibrium method and its rela- Phys. Solids 12, 337-351.
tion to slip line method. Soils Fdns 31, 1-13. Terzaghi, K. (1943). Theoretical soil mechanics. New
Famiyesin, 0. 0. (1990). Modelling and computational York: Wiley.
aspects of the nonlinear finite element analysis of Vermeer, P. A. (1990). The orientation of shear bands in
general concrete structures. University of Swansea, biaxial tests. GCotechnique 40, No. 2,223-236.
PhD thesis. Vermeer, P. A. & de Borst, R. (1984). Non-associated
Goldscheider, M. (1988). Private communication. plasticity for soils, concrete and rock. Heron 29,
Grifflths, D. V. (1982). Computation of bearing capacity 1-64.
factors using finite elements. Ghotechnique 32, No. 3, Zienkiewicz, 0. C., Humpheson, C. & Lewis, R. W.
195-202. (1975). Associated and non-associated visco-
Hill, R. (1956). The mathematical theory of plasticity. plasticity and plasticity in soil mechanics.
Oxford: Clarendon. GCotechnique 25, No. 4, 671-689.